Speed up ReID

[migsdigs]

Hi there! Thank you for your very nice work.

I was wondering if it is possible to increase the speed of the ReID, as it appears to be the bottleneck. Running the mc_demo_yolov7.py for example, I am able to run without reid and with sparseOptFlow as the CMC and get around 30 FPS (real time), enabling the reid, it drops as low as 10 FPS. Is all of the reid computation done on the GPU?

I would appreciate any advice as to how to improve this.

Many thanks